Key Takeaways
OpenAI announced improved plugin ranking and recommendations at DevDay on 29 September 2026
ChatGPT surfaces relevant plugins in its directory and inside conversations
OpenAI puts ChatGPT's reach at 1.2B weekly users
The ranking upgrade, plugin extensions and Plugin Creator are available on all plans
OpenAI's plugin docs name metadata as what ChatGPT uses to pick a tool
ChatGPT plugins now get recommended while people chat.
At DevDay on 29 September, OpenAI said improved ranking and recommendations now help people find plugins in the directory and in conversations.
The change got 1 sentence in a recap of more than 20 announcements. For your brand, it's the one to read twice.
OpenAI says developers can now launch experiences to 1.2B weekly users. A plugin that ChatGPT suggests mid-chat reaches a person at the moment they ask.
What did OpenAI change about ChatGPT plugins?
OpenAI's DevDay 2026 recap, published on 29 September 2026, says improved ranking and recommendations help people find ChatGPT plugins in the directory and in conversations. OpenAI also shipped Plugin Creator and a new submission flow on all plans.
A suggestion is still only an offer. The person in the chat has to accept it.
"Users choose which plugins to use and approve the access each one receives." - OpenAI, DevDay 2026 Recap
If you already track how ChatGPT recommends products to shoppers, plugins are the next surface to watch.
4 DevDay changes that affect plugin discovery
- Ranking and recommendations. ChatGPT surfaces relevant plugins in the directory and in conversations.
- Plugin extensions. A plugin can hold a sidebar home, interactive panels and file viewers inside ChatGPT.
- Shareable profiles. People can showcase Sites and plugins that others find and reuse.
- Sign in with ChatGPT. People can sign in to partner tools with a ChatGPT account and connect plugins faster.
Who gets each plugin feature
| DevDay change | What it does | Who gets it |
|---|---|---|
| Ranking and recommendations | Surfaces plugins in the directory and in chats | All plans |
| Plugin extensions | Sidebar home, panels and file viewers | All plans |
| Plugin Creator and new submission flow | Faster builds and clearer review feedback | All plans |
| MCP events | Plugins start automations from app events | All plans |
| Sites can host plugins | Teammates use a plugin with their own data | Business, Enterprise, Healthcare, Edu |
| Shareable profiles | Public showcase of Sites and plugins | Free, Go, Plus, Pro, Business, Enterprise |
| Sign in with ChatGPT | 1 account across partner tools | Identity is global |
The discovery features are on all plans, so your plugin competes in everyone's chats.
Metadata is the 1 lever OpenAI confirms
OpenAI's plugin docs say ChatGPT decides when to call your tool based on the metadata you provide. That covers tool names, descriptions and parameter docs.
The same page tells you to test 3 prompt types. Direct prompts name your product. Indirect prompts describe the outcome a person wants.
Negative prompts are cases another tool should handle. Indirect prompts are where discovery happens, because the person never types your name.
The DevDay recap gives no detail on ranking signals. Treat metadata as the only confirmed input for now.
3 things to do this week
Rewrite each tool description so it opens with the situation it's for. OpenAI's docs recommend that pattern.
List the indirect prompts your customers would type. Then check how often ChatGPT calls your plugin for each one.
Keep an eye on the paid side as well. OpenAI already runs a crawler that checks ad landing pages.
Disclaimer: This article is AI-assisted and may contain errors, so please verify key details in your own research.
.png&w=3840&q=75)
